About SDH-Engine
How this drug repurposing engine works, what the scores mean, and why it exists
SDH-Engine was built by a patient diagnosed with SDH-deficient GIST (gastrointestinal stromal tumor). SDH-deficient cancers are rare. Information is scattered across journals, databases, and clinical trial registries. There is no dedicated cure.
This tool exists to systematically map the molecular consequences of SDH deficiency to existing approved drugs that could potentially be repurposed — accelerating the path from biology to treatment by connecting dots that are hard to connect manually.
The entire application was built using Claude Code(Anthropic's AI coding tool) in collaboration with the patient. The AI layer that powers drug scoring and hypothesis generation runs on Claude by Anthropic.
SDH-Engine maps the molecular cascade from SDH loss to druggable targets in three layers:
A curated knowledge base encoding the complete SDH deficiency molecular cascade — from succinate accumulation through 7 downstream druggable pathways, 16 molecular targets, and their biological relationships.
Claude AI analyzes drug candidates against the SDH biology context, scores evidence across multiple dimensions, generates novel repurposing hypotheses, and summarizes research papers in plain language.
Live data from PubMed (research papers), ClinicalTrials.gov (active trials), OpenTargets (drug-target associations), and ChEMBL (bioactivity data) enriches the curated knowledge base with real-world evidence.
Succinate dehydrogenase (SDH), also known as mitochondrial Complex II, is an enzyme that sits at the intersection of the TCA cycle and the electron transport chain. It has four subunits: SDHA, SDHB, SDHC, SDHD, plus assembly factors (SDHAF1, SDHAF2).
Loss-of-function mutations in any subunit cause succinate to accumulate massively (100x+ normal levels). Succinate acts as an oncometabolite, triggering a cascade of downstream effects:
Pseudohypoxia / HIF
Succinate inhibits PHD enzymes, stabilizing HIF-1α/2α regardless of oxygen. This drives VEGF, glycolysis, and angiogenesis.
Epigenetic Dysregulation
Succinate inhibits TET DNA demethylases and KDM histone demethylases, causing global hypermethylation and tumor suppressor silencing.
VEGF Signaling
Downstream of HIF activation, VEGF/VEGFR2 signaling drives tumor angiogenesis — new blood vessels feeding the tumor.
mTOR / PI3K / AKT
Metabolic reprogramming activates the PI3K/AKT/mTOR growth signaling axis through multiple upstream inputs.
Glutamine Dependency
With the TCA cycle broken at Complex II, cells become addicted to glutamine for fuel via reductive carboxylation.
Oxidative Stress / ROS
Complex II dysfunction causes electron leak, increasing reactive oxygen species — a vulnerability that could be therapeutically exploited.
Autophagy / Survival
Metabolic stress triggers autophagy as a survival mechanism, which itself becomes a potential therapeutic target.
Each pathway represents a potential therapeutic intervention point. The drug candidates in this engine target one or more of these pathways.
SDH mutations cause several distinct tumor types, collectively known as SDH-deficient tumor syndrome:
GIST
Gastrointestinal Stromal Tumor
5-7.5% of all GISTs are SDH-deficient. Unlike KIT/PDGFRA-mutant GIST, these are resistant to imatinib.
Paraganglioma
Extra-adrenal Neuroendocrine Tumor
Especially associated with SDHB mutations. Can be metastatic.
Pheochromocytoma
Adrenal Medullary Tumor
Associated with SDHB and SDHD mutations. Produces excess catecholamines.
Renal Cell Carcinoma
SDH-deficient RCC
A distinct WHO-recognized subtype of kidney cancer.
Pituitary Adenoma
Pituitary Tumor
Rare, associated with SDHA and SDHB mutations.
Each drug candidate receives an evidence score (0-100) based on a weighted evaluation across five dimensions. The AI scoring uses Claude to assess each dimension, combining quantitative evidence counting with qualitative mechanistic reasoning.
How directly does this drug target the SDH-loss molecular cascade? Is the mechanism of action clearly connected to one of the 7 downstream pathways?
Is there in vitro or in vivo data showing activity in SDH-deficient models? Have cell lines or animal models been tested?
Are there case reports, case series, or clinical trials showing response in SDH-deficient patients? This is the strongest form of evidence.
Is the evidence specifically about SDH-deficient tumors, or extrapolated from general cancer data? SDH-specific evidence is weighted more heavily.
Is the drug FDA-approved? Orally available? Well-tolerated for long-term use? Accessible to patients? Practical factors that affect real-world repurposing feasibility.
Score interpretation:
Drugs with clinical use or strong clinical evidence in SDH-deficient or closely related tumors. These are currently being used or actively studied in relevant patient populations.
Drugs currently in clinical trials for SDH-deficient tumors or with trial data in closely related conditions (e.g., paraganglioma, VHL disease).
Drugs with preclinical evidence (cell lines, animal models) suggesting activity against SDH-deficient biology, but without clinical trial data yet.
Drugs with a strong mechanistic rationale for targeting SDH-deficient pathways, but limited or no direct experimental evidence. These represent hypotheses worth investigating.
The hypothesis generator uses Claude Opus (Anthropic's most capable model) with adaptive thinking to propose novel drug repurposing candidates not already in the database.
The AI receives:
- The complete SDH biology context (~2,500 tokens of curated molecular oncology knowledge)
- All current drug candidates and their pathway mappings
- The selected focus pathway (if any)
- User-provided constraints or context
It then reasons about synthetic lethal interactions, metabolic vulnerabilities, cross-disease parallels (IDH-mutant, FH-deficient, VHL), and combination strategies to propose 3-5 novel candidates with mechanistic rationale and confidence levels.
Important:AI-generated hypotheses are starting points for investigation, not clinical recommendations. Always discuss any treatment ideas with your oncology team. The confidence levels reflect the AI's assessment of the mechanistic plausibility, not clinical proof.
PubMed / NCBI
Biomedical literature database. Used for the research feed — searching and displaying published papers on SDH-deficient diseases.
ClinicalTrials.gov
U.S. National Library of Medicine database of clinical studies. Used to find active and completed trials relevant to SDH-deficient cancers.
Open Targets
Drug-target-disease association platform. Used for evidence enrichment — connecting drugs to targets and diseases with association scores.
ChEMBL
Bioactivity database from the European Bioinformatics Institute. Used for compound details, mechanisms of action, and bioactivity data.
Claude AI (Anthropic)
Powers the AI scoring, hypothesis generation, and paper summarization. The biology context is curated and validated, not AI-generated.
Disclaimer: SDH-Engine is a research and information tool built by a patient, not a medical device. The drug candidates, evidence scores, and AI-generated hypotheses are intended for research exploration and discussion with your medical team — not as clinical recommendations. Always consult your oncologist or healthcare provider before considering any treatment changes. The scoring reflects available evidence and mechanistic plausibility, not clinical efficacy.